Abstract

Underwater visual perception is severely impeded by optical degradation such as wavelength-dependent light absorption, scattering, and contrast attenuation. While substantial research has focused on developing image restoration and enhancement algorithms as preprocessing steps, the question remains whether explicit image enhancement is truly a prerequisite for robust downstream object detection. This study presents a systematic and critical evaluation assessing the necessity and efficacy of image enhancement in underwater object detection pipelines. By coupling state-of-the-art underwater enhancement methods with leading object detection architectures across extensive aquatic benchmarks, we analyze whether visual quality metrics correlate with downstream detection precision. The empirical findings reveal that while enhancement algorithms improve human visual appeal, they do not uniformly boost detection accuracy and may even introduce artifacts that degrade localization performance. Based on these insights, we discuss task-driven alternatives and lightweight end-to-end architectures tailored for real-world underwater robotic systems.